Carcinogenesis vol.35 no.9 pp.2047–2054, 2014 doi:10.1093/carcin/bgu098 Advance Access publication April 29, 2014

Aromatic adducts and lung cancer risk in the European Prospective Investigation into Cancer and Nutrition (EPIC) Spanish cohort Tamra Gilberson1, Marco E.M.Peluso2, Armelle Munia2, Leila Luján-Barroso1, María-José Sánchez3,4, Carmen Navarro4,5, Pilar Amiano4,6, Aurelio Barricarte4,7, J.Ramón Quirós8, Esther Molina-Montes3,4, Emilio SánchezCantalejo3,4, María-José Tormo4,5, María-Dolores Chirlaque4,5, Eva Ardanaz4,7, Miren Dorronsoro4,6, Massimo Confortini9, Catalina Bonet1, Núria Sala1,10, Carlos A.González1 and Antonio Agudo1,* 1

*To whom correspondence should be addressed. Tel: +34 932607401 ext. 3075; Fax: +34 932607787; Email: [email protected]

In this case-cohort study, we examined the association between bulky DNA adducts and the risk of lung cancer within the European Prospective Investigation into Cancer and Nutrition (EPIC) Spanish cohort with an average 7-year follow-up, including 98 cases of primary lung cancer and 296 subjects randomly selected from the cohort. Aromatic adducts were measured using 32 P-postlabeling in leukocyte DNA from blood samples collected at enrollment. The association between DNA adducts and the risk of lung cancer was estimated using a Cox proportional hazards model with a modified partial likelihood. There was an overall significant increased risk for developing lung cancer when DNA adduct concentrations were doubled, with relative risk (RR) adjusting for all relevant confounders of 1.36 with 95% confidence interval (CI) 1.18–157. There was a significant increased risk for developing lung cancer when DNA adduct concentrations were doubled for current smokers and among subjects exposed to PAH at work; there was also a slightly higher increase among males than females. However, no statistically significant differences were observed for the effect of adduct levels across smoking status, sex or occupational exposure to PAH. A meta-analysis combined four prospective studies, including this study, resulting in a significant association among current smokers, with an overall estimate of 34% increase in the risk of lung cancer when doubling the level of aromatic DNA adducts in leukocytes.

Introduction Worldwide, lung cancer is the leading cause of cancer mortality in men and the second in women. Further, an estimated 1.6 million new lung cancer cases and 1.4 million deaths were expected to occur globally by 2008 (1). Smoking tobacco is the principle cause of lung Abbreviations: BaP, benzo(a)pyrene; BMI, body mass index; CI, confidence interval; EPIC, European Prospective Investigation into Cancer and Nutrition; PAH, polycyclic aromatic hydrocarbons; RR, relative risk; WBC, white blood cells.

Materials and methods Design and study population This investigation was conducted using a case-cohort approach among 41 438 individuals from the EPIC Spanish cohort. Male and female subjects between the ages of 29 and 69  years were recruited between 1992 and 1996 in five

© The Author 2014. Published by Oxford University Press. All rights reserved. For Permissions, please email: [email protected]

2047

Downloaded from http://carcin.oxfordjournals.org/ at University of Michigan on July 12, 2015

Unit of Nutrition, Environment and Cancer, Cancer Epidemiology Research Program, Catalan Institute of Oncology-IDIBELL, L’Hospitalet de Llobregat, Barcelona 08908, Spain, 2Cancer Risk Factor Branch, ISPOCancer Prevention and Research Institute, Florence 50131, Italy, 3Granada Cancer Registry, Andalusian School of Public Health, Granada 18080, Spain, 4 CIBER de Epidemiología y Salud Pública (CIBERESP), Barcelona, Spain, 5 Department of Epidemiology, Murcia Health Council, Murcia 30003, Spain, 6 Public Health Division of Gipuzkoa, BIODonostia Research Institute, Department of Health of the regional Government of the Basque Country, San Sebastian 48902, Spain, 7Public Health Institute of Navarra, Pamplona 31003, Spain, 8Public Health Directorate, Asturias 31003, Spain, 9Unit of Analytical Citology and Bimolecular, ISPO-Cancer Prevention and Research Institute, Florence 50131, Italy and 10Molecular Epidemiology Group, Translational Research Laboratory, Catalan Institute of Oncology-IDIBELL, L’Hospitalet de Llobregat, Barcelona 08908, Spain

cancer, accounting for nearly 82% of lung cancer cases in European countries including Spain (2). This is probably due to the presence of carcinogens found in tobacco smoke including aromatic compounds such as polycyclic aromatic hydrocarbons (PAH). After tobacco, key sources of exposure to aromatic compounds are found in environmental and occupational air pollution and the human diet (3). Aromatic compounds form metabolites that can covalently bind to DNA and form adducts and when left unrepaired can cause mutations which in turn may lead to altered DNA replication (4). Therefore, DNA adducts are not only biomarkers of exposure, but rather they are integrated markers of DNA damage by aromatic compounds, including metabolic activation of carcinogens and DNA repair mechanisms (4,5). Although the relationship between DNA adduct levels in white blood cells (WBC) and tumors in specific tissue has not been clearly established, several human and experimental studies have demonstrated that WBC act as a suitable surrogate for DNA adduct levels (5,6). Immunoassays and physicochemical methods, such as synchronous scanning fluorescence spectrometry, have been used to detect DNA adducts formed through benzo(a)pyrene (BaP) diolepoxide, the major intermediate in the activation of BaP (7). The 32P-postlabeling technique has been proven to be an effective method to measure DNA adducts due to its high sensitivity and small amounts of DNA required (8). This method does not however allow for the structural identification of aromatic adducts (9). Few studies have explored DNA adducts in WBC as a precursor to lung cancer. Among the epidemiological studies known to us that investigate such relationship, four were case–control studies (10–13), and three used a prospective approach (14–16). Three of the case– control studies reported a higher lung cancer risk among persons with high adduct levels (10,12,13), whereas the other found no association between DNA adduct levels and the risk of lung cancer (11). Of the three prospective studies, a nested case–control design with 89 lung cancer cases reported higher levels of DNA adducts among current smokers associated with a higher risk of lung cancer among US males (14). A  European case–control study nested within the European Prospective Investigation into Cancer and Nutrition (EPIC) cohort restricted to never smokers and long-term quitters included 115 lung cancer cases and reported a significant excess risk of lung cancer associated with higher DNA adduct levels among never smokers (15). A Danish case-cohort study included 245 cases of lung cancer and found a weak association between bulky DNA adducts and lung cancer risk in current smokers (16). A pooled analysis of these three prospective studies reported an augmented risk of lung cancer for increasing levels of aromatic adducts in current smokers (17). This study examines the association between bulky DNA adducts and the risk of lung cancer in the Spanish adult population including both men and women using a prospective case-cohort design. The study further examines DNA adduct levels in categories by smoking status, histology, occupational exposure and dietary BaP as potential pathways of aromatic compounds on the risk of developing lung cancer. Our hypothesis was that the amount of adducts in nuclear DNA in WBC would be higher in the subjects that developed lung cancer compared with the subcohort, possibly due to increased exposure to aromatic compounds such as PAH. We further added our results to the three previous prospective studies and estimated an overall effect of DNA adducts on lung cancer by means of a meta-analytic approach.

T.Gilberson et al.

Baseline information Food and nutrient intake throughout the year, accounting for seasonal variation, was estimated through a personal interview with a validated dietary history questionnaire developed for the EPIC project (21,22). The questionnaire included foods and beverages organized by meals, including regional dishes. Trained staff gathered data on cooking methods, average consumption frequency per week and portion size for foods consumed at least twice per month and once per month for seasonal foods. Data on composition of foods were used to estimate the total caloric intake and the daily intake of macro and micronutrients. The daily intake of BaP was estimated using a compiled database with information on the content in foods of several potential carcinogens (23). Anthropometric measurements including height and weight were taken by trained individuals and used to calculate the body mass index (BMI, kg/m2). A  lifestyle questionnaire included information on smoking habits, occupational history, education, history of previous illnesses and physical activity. Smoking behaviors were registered as smoking status at baseline: never, current and former smokers (of at least 1 year); current smokers were further divided according to number of cigarettes smoked per day (less or more than 20 cigarettes per day), whereas former smokers were classified according to time since quitting (less or more than 10 years). Physical activity was assessed using a validated questionnaire focused on usual activity in occupational, household and recreational (including walking, cycling and sport activities) domains. Information was combined and subjects were classified as inactive, moderately inactive, moderately active and active (24). The lifestyle questionnaire included several questions on occupational history of the participants focusing on 52 selected jobs that have been previously linked to the risk of developing cancer, and these occupations were then combined according to presumed common carcinogenic exposures. Following an earlier analysis on occupational exposures and lung cancer incidence in the EPIC cohorts (25), the following carcinogenic exposures were considered relevant to lung cancer risk: asbestos, heavy metals, PAH, environmental tobacco smoke and silica. Another variable was created by categorizing possible exposure to PAH at work from a list of occupations compiled by the authors including farming, mining, transport, bus or taxi driver, truck driver, asphalt worker, car mechanic, roofing, foundries, rubber industry, metal, construction, bartender, restaurant, welding, steel mill, demolition and chemical industry. DNA adducts determination Blood samples were analyzed at the ISPO-Cancer Prevention and Research Institute in Florence, Italy, using 0.5-ml aliquots of buffy coat stored in liquid nitrogen at −196°C. Exactly the same procedures for DNA extraction and adducts measurement were applied to referents (subcohort) and lung cancer cases, but there was no randomization of samples to the lab batches. Due to logistical reasons, the assays were not done at the same time; first were processed samples from the subcohort members and afterwards those from the cases. WBC DNA from the subcohort’s subjects and the cases was extracted and purified using a method that requires digestion with ribonuclease A, ribonuclease T1 and proteinase K treatment and extraction with saturated phenol, phenol/chloroform/isoamyl alcohol (25:24:1), chloroform/isoamyl alcohol (24:1) and ethanol precipitation (26). DNA concentration and purity were determined using a spectrophotometer. Coded DNA samples were subsequently stored at −80°C until laboratory analyses were realized. Aromatic DNA adducts were

2048

analyzed blindly using the nuclease P1 modification of the 32P-DNA postlabeling technique (27,28). Coded DNA samples (2  μg) were hydrolyzed by incubation with micrococcal nuclease (21.45 mU/μl) and spleen phosphodiesterase (6.0 mU/μl) in 5.0 mM Na succinate, 2.5 mM calcium chloride, pH 6.0 at 37°C for 4.5 h. Hydrolyzed samples were treated with nuclease P1 (0.1 U/μl) in 46.6 mM sodium acetate, pH 5.0, and 0.24 mM ZnCl2 at 37°C for 30 min (28). After enzymatic treatment, 1.8 μl of 0.16 mM Tris base was added to the sample. The nuclease P1 resistant nucleotides were incubated with 12  μCi of carrier-free [γ-32P]ATP (3000 Ci/mM) and polynucleotide kinase T4 (0.75 U/μl) to generate 32P-labeled DNA adducts in bicine buffer, 20 mM bicine, 10 mM MgCl2, 10 mM dithiotreithol, 0.5 mM spermidine, pH 9.0, at 37°C for 30 min (28). Detection of bulky DNA adducts was carried out by polyethyleneimine cellulose thin layer chromatography plates (MachereyNagel, Germany). 32P-labeled samples were applied on different polyethyleneimine cellulose thin layer chromatography plates to analyze the formation of bulky DNA adducts. The chromatographic resolution of bulky DNA adducts was carried out using a high-urea solvent system capable of ensuring efficient detection of aromatic/hydrophobic DNA adducts, such as BaP-related DNA adducts (28): 1.0 M sodium phosphate, pH 6.8 for the preparatory chromatography; 4.0 M lithium formate; 7.5 M urea, pH 3.5 and 0.65 M lithium chloride, 0.45 M Tris base, 7.7 M urea, pH 8.0 for the two dimensional chromatography; and 1.7 M sodium phosphate pH 5.0 for the clean-up chromatography. The levels of aromatic DNA adducts were expressed as relative adduct labeling = pixels in adducted nucleotides/pixels in total nucleotides. The values of DNA adducts were corrected across experiments based on the recovery of BaP DNA adduct standard (kindly donated by Prof. F. A. Beland, National Center for Toxicological Research, Jefferson, AR) that was analyzed in parallel. The detection limit was 0.1 adduct per 109 normal nucleotide (29), and all analyses were carried out blindly.

Statistical methods DNA adduct levels were expressed as adducted nucleotides per 109 normal nucleotides. Samples below the detection limit (0.1 × 10−9) were assigned a value of 0.05 × 10−9. Given the right-skewed distribution of the DNA adduct levels, the data were log transformed to stabilize the variance and normalize the distribution. Geometric means (adjusted by age, sex, center and season) are provided to describe the DNA adduct levels, and a t-test was used to compare the DNA adduct levels in cases and the subcohort. The association between DNA adducts and lung cancer was evaluated by estimating the relative risk (RR) and 95% confidence intervals (CI) based on the hazard ratios from a proportional hazards model (Cox regression). A modified partial likelihood (30) was used to take into account the case-cohort design, applying the weighting method proposed by Prentice (31), as it seems to provide the estimates that most resemble those of the full cohort. Age entry and age exit were used for time scale in all Cox models. Adduct levels were evaluated using both a categorical variable based on tertiles in the subcohort and the continuous variable using the base 2 logarithm (log2) of adduct concentration. The exponential of the parameter estimated for this variable may be interpreted as the RR of lung cancer associated with a doubling in the level of DNA adducts. Four models of increasing complexity were built using Cox regression. The minimally adjusted model included sex, age and center to take into account the stratified design of the subcohort, as well as season of blood extraction, found to be associated with adduct levels in the subcohort in previous analyses (19). A second model built upon the minimally adjusted model to account for the possible effects of confounders other than those assumed to be potential sources of PAH exposure (model 2). Confounders included education, physical activity, BMI, total energy intake, occupational exposure with potential risk for lung cancer and consumption of fruits, vegetables and vitamin B6. Next, three models were built into model 2 with separate variables that represented potential sources of exposure to PAH: smoking exposure (model 3a) using smoking status, cessation and intensity (never smokers; former smokers, greater or less than 10 years since quitting; and current smokers, greater or less than 20 cigarettes per day); dietary intake of BaP (model 3b) and occupational exposure to PAH (model 3c). Finally, the fully adjusted model (model 4) included all design variables, potential confounders and potential sources of PAH: age, sex, center, season, education, BMI, physical activity, smoking status, work in occupations with potential risk of lung cancer, occupational exposure to PAH and dietary variables including total energy intake, dietary BaP, vegetables, fruit and vitamin B6. Subgroup analysis was carried out for sex, smoking status, dietary BaP and occupational exposure to PAH. Basic smoking status was considered as never, former and current smokers. High and low dietary BaP intake was based upon the median of BaP (0.12 µg per day) in the EPIC-Spain cohort (19). The potential interaction between these factors and the adduct levels on the risk of lung

Downloaded from http://carcin.oxfordjournals.org/ at University of Michigan on July 12, 2015

Spanish provinces: Asturias, Gipuzkoa, Navarra, Murcia and Granada. The cohort was contacted in health centers and blood donation centers varying between regions. At recruitment, information on diet and lifestyle factors, anthropometrical measurements and 30 ml blood samples were collected. Further details about the EPIC cohort can be found elsewhere (18). The subcohort consisted of 300 individuals (60 from each of the five centers) randomly selected by stratified sampling according to the age–sex structure of the Spanish population. Biological material was not available for four participants leaving 296 individuals (147 men and 149 women) in the final analysis (19). The cases were subjects of the full Spanish EPIC cohort defined as first occurrence of a primary tumor of the lung (ICD-O-2 code C34). Identification of cases was done by record linkage of the cohort databases with population-based tumor registries from the regions where the cohort groups were recruited. In addition, information on vital statistics and causes of death were obtained by crossing data with the National Institute of Statistics (INE). In total, 99 cases of lung cancer were reported by the end of monitoring in 2001 with an average 7-year follow-up. Due to lack of distinctive biological material from one individual, final analysis was carried out with 98 cases. Histological data were categorized based on morphology codes grouped in categories according to the WHO scheme (20). Data regarding histological type for two of the subjects were missing. One of the lung cancer cases was previously selected as a member of the subcohort.

Aromatic adducts and lung cancer

cancer was assessed by comparing the likelihood of the model with the corresponding interaction terms and the model containing terms for the main effects only by means of the likelihood ratio test. Subgroup analysis was also carried out for the main histological types. Finally, a meta-analysis was carried out including the results of the three prospective studies mentioned above (14–16) and results from this study. Because the baseline level of adducts varied markedly across studies, effect estimates according to categories of exposure were not directly comparable. However, all of the studies provided estimates of the RR by log-transformed (natural log) DNA adducts levels. We calculated the RR (and 95% CI) for the log2 adducts using the formula exp[ln(RR) × ln(2)]. Overall RRs were estimated by a meta-analytic approach using a random effect model (32). Heterogeneity across studies was assessed by means of the Cochran’s Q and the I2 statistic (33). Because the various studies were heterogeneous in regards to smoking habits of the underlying populations, the meta-analysis was stratified by smoking status. All statistical tests were two-sided, and P values less than 0.05 were considered to be statistically significant. STATA 10.0 was used for all analyses (34).

Baseline characteristics comparing the 98 lung cancer cases with the 296 subcohort subjects are outlined in Table I. Cases were slightly older than referents (subcohort), with a mean age of 52.2 and 49.1  years, respectively, and were predominantly males (81% among cases versus 50% in the subcohort). Cases tended to have a lower level of education, a significantly higher proportion of heavy smokers (current smokers of 20 or more cigarettes per day), occupational exposures to PAH or potential employment involving lung cancer risk, whereas no differential patterns of BMI and physical activity were observed between both groups. Cases had significantly higher energy intake and lower consumption of fruit and total carotenoids. Almost one third (31%) of lung cancer cases were adenocarcinomas, and one out of five (21%) were squamous cell carcinomas, with a proportion of cases around 16% for small cell and large cell carcinomas (Table II). Bulky DNA adducts were significantly higher (P value, 0.002) in cases than in the subcohort, with geometric means of 7.80 versus 4.09 adduct per 109 nucleotides, respectively (Table III). Among current smokers, a higher overall level of bulky DNA adducts was seen in cases compared to the subcohort (geometric mean 7.00 versus 3.96 adduct per 109 nucleotides, respectively) with significant statistical association (P value, 0.03). However, among never and former smokers, there was a slightly higher level of bulky DNA adducts in cases compared to the subcohort, but no statistical association could be established. Bulky DNA adduct levels in males were significantly higher in cases than in the subcohort (8.05 versus 4.18 adduct per 109 nucleotides, respectively, P value 0.01). No other significant differences were found in smoking categories between cases and the subcohort in males or females. Table IV outlines the Cox regression models used to measure associations between bulky DNA adducts and the risk of developing lung cancer. After minimally adjusting the model by age, sex, center and season, association between bulky DNA adducts and lung cancer risk compared with the lower tertile showed a significant increase in the second tertile, but a non-significant increase for those in the highest tertile; however, there was a significant increase, with an RR of 1.25 (95% CI 1.11–1.40) for the continuous variable (log2 transformed), indicating a risk of lung cancer increase by 25% when doubling the concentration of aromatic adducts. This pattern of non-monotonically increasing risk by tertiles and a trend toward increased risk in the logscale of DNA adduct levels was consistent in all models. Adjusting for education, dietary and physical activity confounders only modestly increased the association of bulky DNA adducts (log2 scale) on lung cancer risk to a RR of 1.29 (95% CI 1.14–1.45). Further, adjusting for smoking exposure increased bulky DNA adduct levels for lung cancer risk to a RR of 1.37 (95% CI 1.19–1.57), whereas adjusting for dietary BaP and occupational exposure maintained relatively similar results to the adjusted model with RR 1.28 and RR 1.27, respectively. Finally, the fully adjusted model (including all potential confounders

Discussion Results from our study indicate an association between levels of bulky DNA adducts in WBC and the risk of lung cancer. Analysis based upon the log2-tranformed continuous variable indicated an increase in risk of lung cancer by 36% when doubling the concentration of aromatic adducts. Association between bulky DNA adducts and lung cancer risk showed a significant increase in the second tertile (as compared with the lowest), but a non-significant increase for those in the highest tertile. This pattern of non-monotonically increasing risk by tertiles of DNA adduct levels, although reported in previous works (17), is intriguing. It can be a result of chance, but it is also compatible with a levelling-off of the dose–response relationship between adduct levels and risk. This pattern fits quite well the data when the level of exposure is expressed as log-dose. Therefore, using a log-linear model for the relationship between adduct levels and lung cancer risk can be acceptable in spite of the non-monotonically increase in risk observed when analyzed by tertiles. The risk of developing lung cancer based on doubling the concentrations of bulky DNA adducts was higher for current smokers and subjects with occupational exposure to PAH although there was no significant interaction between adduct levels and smoking or occupational PAH exposure. Findings of this investigation are in line with the hypothesis that increased levels of bulky DNA adducts in WBC show an association with lung cancer for individuals with high exposure to PAH. A strength of this study is the prospective design that limits reverse causation bias and reduces time order uncertainty. As far as the authors are aware, this study, the Bak et  al case-cohort study (16) and two nested case–control studies within a cohort (14,15) are the only prospective investigations that have reported findings from the predictive assessment of DNA adducts on the risk of developing lung cancer. A key difference between this study and the Tang et al. study is the inclusion of both men and women, whereas Tang et al. (14) study was restricted to male physicians. The Bak et al. study (16) included

2049

Downloaded from http://carcin.oxfordjournals.org/ at University of Michigan on July 12, 2015

Results

and sources of aromatic compounds) resulted in a RR of 1.36 (95% CI 1.18–1.57) for doubling the concentration of aromatic adducts. Subgroup analyses in Table V indicate bulky DNA adducts with a higher risk of developing lung cancer among current smokers presenting a RR of 1.38 (95% CI 1.04–1.83) when DNA adduct levels were doubled while no significant risk was found in former smokers. Slightly higher increase in risk was found among males than in females, although both sexes showed a significant association between doubling DNA adducts levels and lung cancer risk. Bulky DNA adducts associated with the risk of lung cancer was higher among those exposed to PAH at work resulting in a RR of 1.60 (95% CI 1.17–2.18), whereas the non-exposed group to PAH at work showed no statistical significance with a RR of 1.28. In addition, no differences in risk were observed among subjects with low or high intake of dietary BaP. However, no statistically significant effect modification was observed for the RR of adduct levels (log2 scale) either across smoking status categories or according to sex, occupational exposure to PAH or dietary intake of BaP. Finally, regarding histology, the small number of cases did not allow to compute fully adjusted estimates for some major histological types, such as squamous or small cell types. Doubling bulky DNA adduct levels exhibited a significant increased risk for adenocarcinoma (RR of 1.42, 95% CI 1.02–1.97), as well as for the combination of non-small cell lung cancers (RR 1.37, 95% CI 1.16–1.62). Results of the meta-analysis are shown in Figure 1 including RR for doubling concentrations of DNA adducts from prospective studies according to smoking status. When all studies were combined, a significant association was only observed among current smokers, with an overall estimate of 34% increase in the risk of lung cancer for doubling the levels of aromatic DNA adducts in WBC. In spite of the differential composition of the study population regarding smoking habits, as well as differences in the background levels of adducts, there was no statistically significant heterogeneity across studies.

T.Gilberson et al.

Table I.  Baseline characteristics of the study population Cases (n = 98) N

(%)

(%)

N

(15.3) (30.6) (50.0) (4.1)

118 99 79 —

(39.9) (33.4) (26.7)

Aromatic adducts and lung cancer risk in the European Prospective Investigation into Cancer and Nutrition (EPIC) Spanish cohort.

In this case-cohort study, we examined the association between bulky DNA adducts and the risk of lung cancer within the European Prospective Investiga...
608KB Sizes 0 Downloads 4 Views